Research Scientist (Model Evaluation)

About the Company

Sanas is pioneering the future of human communication. Founded by a team of Stanford researchers and entrepreneurs, Sanas has developed the world's first real-time speech AI platform capable of accent translation, noise cancellation, and speech enhancement. The company is a fast-growing startup in Silicon Valley, backed by leading investors like Google Ventures and Insight Partners.

Responsibilities

• Design and own evaluation frameworks across the full model portfolio, including Accent Translation and Noise Cancellation.
• Develop novel quantitative metrics for subjective qualities like accent similarity and speaker identity preservation.
• Build and maintain automated evaluation pipelines that run continuously against model checkpoints.
• Design and operate human evaluation programs, including listener panels and crowdsourced annotation.
• Partner with ML research and product teams to translate quality questions into measurable protocols.

Requirements

• 4+ years of research or applied research experience in speech, audio, or NLP.
• Deep familiarity with speech and audio quality metrics (MOS, MUSHRA, PESQ, STOI, etc.).
• Strong engineering skills in Python and PyTorch to build production-quality evaluation pipelines.

Preferred Qualifications

• Experience evaluating models across ASR, TTS, or machine translation.
• Background in psychoacoustics or perceptual audio quality.
• Published research at INTERSPEECH, ICASSP, ACL, or EMNLP.